Progress of deep learning in oil and gas reservoir geological modeling

Abstract

With the rapid development of big data and deep learning based on artificial intelligence technology, reservoir geological modeling has gradually moved from traditional two-point geostatistical modeling, object-based modeling, multi-point geostatistical modeling, and sedimentary process-based modeling to intelligent geological modeling stage.Intelligent geological modeling based on deep learning mainly uses adversarial generation networks to build a 3D geological model.At present, these studies focus on the improvement of network architectures and algorithms, especially the conditioning of various types of observed data such as seismic and well logging.Few studies focus on sample data obtaining.At present, most of the training samples used in the research are synthetic data based on object modeling or sedimentary process methods.To truly apply this technology to actual underground oil and gas reservoirs, more attention needs to be paid on the acquisition of real sample data.We believe a general artificial intelligence geological modeler is the main direction in the future.However, it is difficult to achieve technological breakthroughs only by the statistical learning method of deep neural networks.The combination of statistical learning and symbol learning may be the only way to realize this technology.

Keywords

• intelligent geological modeling
• deep learning based geological modeling
• conditional generative adversarial network
• artificial intelligence